drm/i915: Split obj->cache_coherent to track r/w

Another month, another story in the cache coherency saga. This time, we
come to the realisation that i915_gem_object_is_coherent() has been
reporting whether we can read from the target without requiring a cache
invalidate; but we were using it in places for testing whether we could
write into the object without requiring a cache flush. So split the
tracking into two, one to decide before reads, one after writes.

See commit e27ab73d17ef ("drm/i915: Mark CPU cache as dirty on every
transition for CPU writes") for the previous entry in this saga.

v2: Be verbose
v3: Remove unused function (i915_gem_object_is_coherent)
v4: Fix inverted coherency check prior to execbuf (from v2)
v5: Add comment for nasty code where we are optimising on gcc's behalf.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101109
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101555
Testcase: igt/kms_mmap_write_crc
Testcase: igt/kms_pwrite_crc
Signed-off-by: Chris Wilson <[e-mail]>
Cc: Maarten Lankhorst <[e-mail]>
Cc: Dongwon Kim <[e-mail]>
Cc: Matt Roper <[e-mail]>
Cc: Joonas Lahtinen <[e-mail]>
Cc: Mika Kuoppala <[e-mail]>
Tested-by: Maarten Lankhorst <[e-mail]>
Acked-by: Maarten Lankhorst <[e-mail]>
Link: https://patchwork.freedesktop.org/patch/msgid/[e-mail]
Reviewed-by: Joonas Lahtinen <[e-mail]>

drivers/gpu/drm/i915/Makefile

···
 	  i915_gem_gtt.o \
 	  i915_gem_internal.o \
 	  i915_gem.o \
+	  i915_gem_object.o \
 	  i915_gem_render_state.o \
 	  i915_gem_request.o \
 	  i915_gem_shrinker.o \

drivers/gpu/drm/i915/i915_drv.h

···
 		     unsigned long addr, unsigned long pfn, unsigned long size,
 		     struct io_mapping *iomap);
 
-static inline bool i915_gem_object_is_coherent(struct drm_i915_gem_object *obj)
-{
-	return (obj->cache_level != I915_CACHE_NONE ||
-		HAS_LLC(to_i915(obj->base.dev)));
-}
-
 #endif

drivers/gpu/drm/i915/i915_gem.c

···
 	if (obj->cache_dirty)
 		return false;
 
-	if (!obj->cache_coherent)
+	if (!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE))
 		return true;
 
 	return obj->pin_display;
···
 
 	if (needs_clflush &&
 	    (obj->base.read_domains & I915_GEM_DOMAIN_CPU) == 0 &&
-	    !obj->cache_coherent)
+	    !(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
 		drm_clflush_sg(pages);
 
 	__start_cpu_write(obj);
···
 	if (ret)
 		return ret;
 
-	if (obj->cache_coherent || !static_cpu_has(X86_FEATURE_CLFLUSH)) {
+	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
+	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
 		ret = i915_gem_object_set_to_cpu_domain(obj, false);
 		if (ret)
 			goto err_unpin;
···
 	if (ret)
 		return ret;
 
-	if (obj->cache_coherent || !static_cpu_has(X86_FEATURE_CLFLUSH)) {
+	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
+	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
 		ret = i915_gem_object_set_to_cpu_domain(obj, true);
 		if (ret)
 			goto err_unpin;
···
 
 	list_for_each_entry(vma, &obj->vma_list, obj_link)
 		vma->node.color = cache_level;
-	obj->cache_level = cache_level;
-	obj->cache_coherent = i915_gem_object_is_coherent(obj);
+	i915_gem_object_set_cache_coherency(obj, cache_level);
 	obj->cache_dirty = true; /* Always invalidate stale cachelines */
 
 	return 0;
···
 {
 	struct drm_i915_gem_object *obj;
 	struct address_space *mapping;
+	unsigned int cache_level;
 	gfp_t mask;
 	int ret;
 
···
 	obj->base.write_domain = I915_GEM_DOMAIN_CPU;
 	obj->base.read_domains = I915_GEM_DOMAIN_CPU;
 
-	if (HAS_LLC(dev_priv)) {
+	if (HAS_LLC(dev_priv))
 		/* On some devices, we can have the GPU use the LLC (the CPU
 		 * cache) for about a 10% performance improvement
 		 * compared to uncached.  Graphics requests other than
···
 		 * However, we maintain the display planes as UC, and so
 		 * need to rebind when first used as such.
 		 */
-		obj->cache_level = I915_CACHE_LLC;
-	} else
-		obj->cache_level = I915_CACHE_NONE;
+		cache_level = I915_CACHE_LLC;
+	else
+		cache_level = I915_CACHE_NONE;
 
-	obj->cache_coherent = i915_gem_object_is_coherent(obj);
-	obj->cache_dirty = !obj->cache_coherent;
+	i915_gem_object_set_cache_coherency(obj, cache_level);
 
 	trace_i915_gem_object_create(obj);
 

drivers/gpu/drm/i915/i915_gem_clflush.c

···
 	 * snooping behaviour occurs naturally as the result of our domain
 	 * tracking.
 	 */
-	if (!(flags & I915_CLFLUSH_FORCE) && obj->cache_coherent)
+	if (!(flags & I915_CLFLUSH_FORCE) &&
+	    obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ)
 		return false;
 
 	trace_i915_gem_object_clflush(obj);

drivers/gpu/drm/i915/i915_gem_execbuffer.c

···
 			eb->request->capture_list = capture;
 		}
 
-		if (unlikely(obj->cache_dirty && !obj->cache_coherent)) {
+		/*
+		 * If the GPU is not _reading_ through the CPU cache, we need
+		 * to make sure that any writes (both previous GPU writes from
+		 * before a change in snooping levels and normal CPU writes)
+		 * caught in that cache are flushed to main memory.
+		 *
+		 * We want to say
+		 *   obj->cache_dirty &&
+		 *   !(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ)
+		 * but gcc's optimiser doesn't handle that as well and emits
+		 * two jumps instead of one. Maybe one day...
+		 */
+		if (unlikely(obj->cache_dirty & ~obj->cache_coherent)) {
 			if (i915_gem_clflush_object(obj, 0))
 				entry->flags &= ~EXEC_OBJECT_ASYNC;
 		}

drivers/gpu/drm/i915/i915_gem_internal.c

···
 				phys_addr_t size)
 {
 	struct drm_i915_gem_object *obj;
+	unsigned int cache_level;
 
 	GEM_BUG_ON(!size);
 	GEM_BUG_ON(!IS_ALIGNED(size, PAGE_SIZE));
···
 
 	obj->base.read_domains = I915_GEM_DOMAIN_CPU;
 	obj->base.write_domain = I915_GEM_DOMAIN_CPU;
-	obj->cache_level = HAS_LLC(i915) ? I915_CACHE_LLC : I915_CACHE_NONE;
-	obj->cache_coherent = i915_gem_object_is_coherent(obj);
-	obj->cache_dirty = !obj->cache_coherent;
+
+	cache_level = HAS_LLC(i915) ? I915_CACHE_LLC : I915_CACHE_NONE;
+	i915_gem_object_set_cache_coherency(obj, cache_level);
 
 	return obj;
 }

drivers/gpu/drm/i915/i915_gem_object.c

···
+/*
+ * Copyright © 2017 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#include "i915_drv.h"
+#include "i915_gem_object.h"
+
+/**
+ * Mark up the object's coherency levels for a given cache_level
+ * @obj: #drm_i915_gem_object
+ * @cache_level: cache level
+ */
+void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
+					 unsigned int cache_level)
+{
+	obj->cache_level = cache_level;
+
+	if (cache_level != I915_CACHE_NONE)
+		obj->cache_coherent = (I915_BO_CACHE_COHERENT_FOR_READ |
+				       I915_BO_CACHE_COHERENT_FOR_WRITE);
+	else if (HAS_LLC(to_i915(obj->base.dev)))
+		obj->cache_coherent = I915_BO_CACHE_COHERENT_FOR_READ;
+	else
+		obj->cache_coherent = 0;
+
+	obj->cache_dirty =
+		!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE);
+}

drivers/gpu/drm/i915/i915_gem_object.h

···
 
 #include <drm/i915_drm.h>
 
+#include "i915_gem_request.h"
 #include "i915_selftest.h"
+
+struct drm_i915_gem_object;
 
 struct drm_i915_gem_object_ops {
 	unsigned int flags;
···
 	 */
 	unsigned long gt_ro:1;
 	unsigned int cache_level:3;
+	unsigned int cache_coherent:2;
+#define I915_BO_CACHE_COHERENT_FOR_READ BIT(0)
+#define I915_BO_CACHE_COHERENT_FOR_WRITE BIT(1)
 	unsigned int cache_dirty:1;
-	unsigned int cache_coherent:1;
 
 	atomic_t frontbuffer_bits;
 	unsigned int frontbuffer_ggtt_origin; /* write once */
···
 	return engine;
 }
 
+void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
+					 unsigned int cache_level);
 void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj);
 
 #endif

drivers/gpu/drm/i915/i915_gem_stolen.c

···
 			       struct drm_mm_node *stolen)
 {
 	struct drm_i915_gem_object *obj;
+	unsigned int cache_level;
 
 	obj = i915_gem_object_alloc(dev_priv);
 	if (obj == NULL)
···
 
 	obj->stolen = stolen;
 	obj->base.read_domains = I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT;
-	obj->cache_level = HAS_LLC(dev_priv) ? I915_CACHE_LLC : I915_CACHE_NONE;
-	obj->cache_coherent = true; /* assumptions! more like cache_oblivious */
+	cache_level = HAS_LLC(dev_priv) ? I915_CACHE_LLC : I915_CACHE_NONE;
+	i915_gem_object_set_cache_coherency(obj, cache_level);
 
 	if (i915_gem_object_pin_pages(obj))
 		goto cleanup;

drivers/gpu/drm/i915/i915_gem_userptr.c

···
 	i915_gem_object_init(obj, &i915_gem_userptr_ops);
 	obj->base.read_domains = I915_GEM_DOMAIN_CPU;
 	obj->base.write_domain = I915_GEM_DOMAIN_CPU;
-	obj->cache_level = I915_CACHE_LLC;
-	obj->cache_coherent = i915_gem_object_is_coherent(obj);
-	obj->cache_dirty = !obj->cache_coherent;
+	i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC);
 
 	obj->userptr.ptr = args->user_ptr;
 	obj->userptr.read_only = !!(args->flags & I915_USERPTR_READ_ONLY);

drivers/gpu/drm/i915/selftests/huge_gem_object.c

···
 		dma_addr_t dma_size)
 {
 	struct drm_i915_gem_object *obj;
+	unsigned int cache_level;
 
 	GEM_BUG_ON(!phys_size || phys_size > dma_size);
 	GEM_BUG_ON(!IS_ALIGNED(phys_size, PAGE_SIZE));
···
 
 	obj->base.read_domains = I915_GEM_DOMAIN_CPU;
 	obj->base.write_domain = I915_GEM_DOMAIN_CPU;
-	obj->cache_level = HAS_LLC(i915) ? I915_CACHE_LLC : I915_CACHE_NONE;
-	obj->cache_coherent = i915_gem_object_is_coherent(obj);
-	obj->cache_dirty = !obj->cache_coherent;
+	cache_level = HAS_LLC(i915) ? I915_CACHE_LLC : I915_CACHE_NONE;
+	i915_gem_object_set_cache_coherency(obj, cache_level);
 	obj->scratch = phys_size;
 
 	return obj;